Thats the way to go. Distill it first, add humans and make it semi-automated, we crack it close to 85% or more.

I think we can talk on this further, we have a team setup and an online interface just doing the same, may be can explore synergies?

You are right in the ways that Daniel says. Our product is more about creating something that *enables* reputation monitoring rather than claiming that we do it all (as you and Daniel say, it’s a very hard problem, though one we are interested in).

One enhancement that is going to come soon is more integration with human-determined sentiment analysis.

Will

Thanks for your comment. You make a compelling argument. Online reputation is most definitely publicity that is either positive, negative, or somewhat neutral.

I think it would be best if we all admit that, technologically speaking, we are nowhere near the capacity to build a sucessful application that can make accurate value judgments.

Obviously, search engines like Google are responsible for making value judgments and delivering relevant search results.

Creating a reputation monitoring application with value judgment capability is even more complicated. The English language is so complicated, that negative adjectives can be positive based on the context of the sentence. And how would an application understand sarcasm? Accurately judging the positive or negative slant of text is simply too complicated at this point in time.

Applications like Google Alert and Distilled’s reputation monitoring service are providing the best one can expect. They collect references to targeted names and terms.

There’s no denying the fact that human discrimination is still required for successful reputation monitoring.

It’s not just a tough nut to crack. It’s nearly impossible. Millions of dollars, expert linguists, and Google-like algorithm experts would be necessary to even attempt an application of this magnitude. Sure, someone will undoubtedly come out with an application, claiming it can accomplish these tasks.

The truth is that a successful application will be on the verge of sentient artificial intelligence before it can accurately discriminate value judgments from linguistic styles.

For now, the industry requires human experts to make value judgments on each public mention of a company name. Besides, someone still needs to expertly strategize a response to each public mention. Just knowing that a libelous statement has been made is not enough. It’s the content, method, and timing of the response that actually turns the potential crisis into positive publicity.

Was impressed with Will’s Distilled and went thru the site, looked at the screenshots, here is what i have to say:
1. Reputation is different from information and what Distilled is doing is providing generic information.
2. To me Reputation means – good, bad or neutral about me/my service/product/features of me/service/product
3. Unless there is a judgment on the talk on my reputation, its not reputation monitoring service.

While Distilled is giving a value-able service, to weed out junk and keep only Relevant results on you, there is no way i can call it reputation monitoring service,

We have been trying to crack it for the past 2 quarters so I know its not an easy nut to find reputation/sentiment thru software.

Thoughts?
Venkat.